Dual-camera refocus works by estimating scene depth, not by truly changing the optical focus after the fact.

With two cameras spaced slightly apart, each sees the scene from a different viewpoint. Software compares the two images to measure parallax—the small shift in object position between the views. From that it builds a depth map, where each pixel is assigned an approximate distance from the camera.

Once the phone has both the normal photo and a depth map, it can simulate shallow depth of field in software. If you tap a subject later, the software treats that depth as the “focus plane” and applies more blur to areas that are farther in front of or behind it, creating an artificial refocus effect.

So the phone usually captures one image and uses depth information to add selective blur afterward. It is not recovering detail that was optically out of focus in the original image; it is emulating the look of a different focus/aperture choice.